459
Recent enhancements to real-time probabilistic thunderstorm guidance products from a time-lagged ensemble of High Resolution Rapid Refresh (HRRR) forecasts

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011
Recent enhancements to real-time probabilistic thunderstorm guidance products from a time-lagged ensemble of High Resolution Rapid Refresh (HRRR) forecasts
Curtis R. Alexander, NOAA/ESRL/GSD and CIRES/Univ. of Colorado, Boulder, CO; and D. A. Koch, S. S. Weygandt, T. G. Smirnova, S. G. Benjamin, and E. P. James

Starting in June of 2009, ESRL GSD began creating a real-time experimental probabilistic thunderstorm guidance product based on the High Resolution Rapid Refresh (HRRR). The HRRR is an hourly updating, convection resolving model. The HRRR utilizes a 3-km horizontal grid spacing configuration of the Weather Research and Forecasting (WRF) model, with a diabatic digital filter initialization (DDFI) radar reflectivity assimilation procedure applied in the parent Rapid Update Cycle (RUC) model. For the 2010 convective season the HRRR domain was expanded to cover all of the U.S. as a further demonstration / evaluation of its utility in providing guidance for convective storms and other mesoscale applications.

The real-time probabilistic thunderstorm guidance product is known as the HRRR Convective Probability Forecast (HCPF) and uses time-lagged ensemble output from the HRRR to create thunderstorm likelihood forecasts. Verification and evaluation of an initial prototype HCPF product over the summer 2009 and 2010 convective seasons has lead to a number of refinements to the algorithm leading to improved performance. These refinements have included: 1) switching from use of the HRRR reflectivity field as the primary predictor to an hourly summed updraft field, 2) upscaling of probabilities to the standard RUC 20 km forecast grid, and 3) implementation of a logistic regression scheme using time/space scaling of a verification field to produce statistically reliable forecast probabilities while retaining forecast resolution and sharpness.

We will describe the HCPF algorithm and illustrate the improvement in skill from the recent enhancements. Presentation of traditional skill score metrics will be augmented by case study examples.